Hi Mohamad,
Thanks for sending. I looked at the images. They seem to be missing large
parts of the brain (see attached), which suggests that the preprocessing
wasn't good, and if that's the case, the results aren't reliable. You may
want to spend some time doing the FSL tutorials and becoming more familiar
with it before embarking on a real project.
Nonetheless, I couldn't replicate your finding of a large cluster. In fact,
there doesn't seem to be signficant results. Please see below (I run a test
with only 500 permutations, but these p-values seem quite far in spite of;
you should use more, say, 5000).
All the best,
Anderson
for x in *corrp* ; do echo "${x} $(fslstats ${x} -R)" ; done
rand_clustere_corrp_tstat1.nii.gz 0.000000 0.328000
rand_clustere_corrp_tstat2.nii.gz 0.000000 0.810000
rand_clustere_corrp_tstat3.nii.gz 0.000000 0.312000
rand_clustere_corrp_tstat4.nii.gz 0.000000 0.394000
rand_clusterm_corrp_tstat1.nii.gz 0.000000 0.398000
rand_clusterm_corrp_tstat2.nii.gz 0.000000 0.824000
rand_clusterm_corrp_tstat3.nii.gz 0.000000 0.342000
rand_clusterm_corrp_tstat4.nii.gz 0.000000 0.418000
rand_tfce_corrp_tstat1.nii.gz 0.000000 0.556000
rand_tfce_corrp_tstat2.nii.gz 0.000000 0.780000
rand_tfce_corrp_tstat3.nii.gz 0.000000 0.226000
rand_tfce_corrp_tstat4.nii.gz 0.000000 0.404000
rand_vox_corrp_tstat1.nii.gz 0.000000 0.706000
rand_vox_corrp_tstat2.nii.gz 0.000000 0.566000
rand_vox_corrp_tstat3.nii.gz 0.000000 0.818000
rand_vox_corrp_tstat4.nii.gz 0.000000 0.552000
On 4 December 2014 at 15:50, Alshikho, Mohamad J. <[log in to unmask]
> wrote:
> Dear Anderson,
> I sent you the images, I hope that you have the chance to look at them.
> Many thanks.
> ________________________________________
> From: FSL - FMRIB's Software Library [[log in to unmask]] on behalf of
> Alshikho, Mohamad J.
> Sent: Monday, December 01, 2014 1:30 PM
> To: [log in to unmask]
> Subject: Re: [FSL] cluster & atlasquery
>
> Dear Anderson,
> Really!! I highly appreciate your interest in my issue. Actually, I am a
> physician and I don't have that great experience in this analysis before.
> This is my first Tbss analysis. I like it and I learned a lot from it as
> from all the guys in this great group.
> My analysis is for two groups: Autism and controls, the age range in this
> study is 5-15 year old.
> The pre-processing steps (bet, ecc, dtifit) ran smoothly. I checked the
> images for all the subjects one by one and the results that I sent to you
> (the images) are consistent with what we are thinking about, it is strongly
> support our hypothesis.
>
> Yesterday I repeated the analysis again running the TBSS scripts from the
> beginning. I changed the threshold in Tbss_4_prestats to 0.3 . In the first
> analysis I used 0.2. What surprised me is the results.
> I got one big cluster for MD and RD while for AD I got the following:
>
> Cluster Index Voxels MAX MAX X (vox) MAX Y (vox) MAX Z (vox) COG X (vox)
> COG Y (vox) COG Z (vox)
> 4 465 0.963 118 90 93 117 99.7 92.2
> 3 186 0.956 123 88 83 124 87.4 82.8
> 2 158 0.955 123 101 98 125 95.8 100
> 1 10 0.95 124 78 84 124 79.9 85.2
>
> for the threshold 0.2 in Tbss_4_prestats I got one big cluster for AD, RD,
> MD
>
> Kindly, is there any relationship between this threshold in
> Tbss_4_prestats and the cluster command? What is the permitted range to
> play with this threshold? Also regarding the global effect. Could you
> please (if possible) refer to me any paper or any link that explain it!
> just to improve my knowledge.
> For the results of one big cluster with many tract. Is it permitted to
> define the coordinates manually in Fslview?
>
> I highly appreciate your time
> With all of my respect
> Mohamad
> ________________________________
> From: FSL - FMRIB's Software Library [[log in to unmask]] on behalf of
> Anderson M. Winkler [[log in to unmask]]
> Sent: Monday, December 01, 2014 6:06 AM
> To: [log in to unmask]
> Subject: Re: [FSL] cluster & atlasquery
>
> Hi Mohamad,
>
> If you inspected the images and they all look fine, and there were no
> errors in the preprocessing, then the result may be legitimate. Or maybe
> your groups are vastly different in ways that they shouldn't probably be --
> something possibly more related to how they were recruited than with the
> data analysis. The command "fslstats" doesn't round as "cluster" and we can
> see that there aren't values equal to 1.
>
> If there is an actual global effect, and it's not interesting for your
> research, you can include a global measure FA (or MD, or AD, or RD,
> respectively) as an extra EV, to treat the global effect as a nuisance,
> thus removing it. This won't fix eventual selection (recruiting) biases,
> though.
>
> I don't think sending me images would help, as the results seem to be
> after all correct given that you verified them. A global effect isn't rare,
> specially
> if the patient group is very different from the control group. In any
> case, you should receive shortly, if not already, a link with instructions
> to upload files.
>
> All the best,
>
> Anderson
>
>
> On 1 December 2014 at 10:06, Niels Bergsland <[log in to unmask]<mailto:
> [log in to unmask]>> wrote:
> Just to add one quick comment, I've always found that cluster outputs
> a maximum of 3 decimal places and thus explains the value of 1 due to
> rounding. Looking at the actual coordinates (e.g. fslview, or matlab),
> will give the precise value.
>
> On Mon, Dec 1, 2014 at 5:21 AM, Alshikho, Mohamad J.
> <[log in to unmask]<mailto:[log in to unmask]>> wrote:
> > Dear Anderson,
> > I highly appreciate your interest in my issue!
> > Sure! I checked all the images. For the subjects one, by one then I
> checked
> > the (slicesdir) from the output of tbss_1_preproc *.nii.gz.
> > The images looks fine.
> > After Randomise I checked the *_Corrp_* images, Also they look fine.
> >
> > I ran fslstats _R for all the images as the following:
> >
> > fslstats tbss_tfce_corrp_tstat1.nii.gz -R
> > 0.000000 0.821300
> > fslstats tbss_tfce_corrp_tstat2.nii.gz -R
> > 0.000000 0.049900
> > slstats tbss_MD_tfce_corrp_tstat1.nii.gz -R
> > 0.000000 0.022100
> > slstats tbss_MD_tfce_corrp_tstat2.nii.gz -R
> > 0.000000 0.998800
> > fslstats tbss_AD_tfce_corrp_tstat1.nii.gz -R
> > 0.000000 0.066900
> > slstats tbss_AD_tfce_corrp_tstat2.nii.gz -R
> > 0.000000 0.965600
> > fslstats tbss_RD_tfce_corrp_tstat1.nii.gz -R
> > 0.000000 0.021600
> > fslstats tbss_RD_tfce_corrp_tstat2.nii.gz -R
> > 0.000000 0.999800
> >
> > Anderson! Kindly, would you mind if I send you some images to your
> personal
> > email, I can't send anything in JISCNAIL.
> >
> >
> > Looking forward to learn from you and again thank you
> >
> > Mohamad
> > ________________________________
> > From: FSL - FMRIB's Software Library [[log in to unmask]<mailto:
> [log in to unmask]>] on behalf of
> > Anderson M. Winkler [[log in to unmask]<mailto:
> [log in to unmask]>]
> > Sent: Sunday, November 30, 2014 1:20 PM
> >
> > To: [log in to unmask]<mailto:[log in to unmask]>
> > Subject: Re: [FSL] cluster & atlasquery
> >
> > Hi Mohamad,
> >
> > Did you open the images to see how do they look like? Do the images seem
> > sensible before running randomise? Did you inspect visually the *_corrp_*
> > images? And did you do so for both contrasts? In the outputs from
> "cluster",
> > is the MAX always 1 for all contrasts, or do you see other values?
> >
> > The command fslstats tells exactly what are the range of values on each
> of
> > the outputs (option -R) and, in addition to visual inspection, it can
> help
> > to diagnose potential issues.
> >
> > All the best,
> >
> > Anderson
> >
> >
> > On 30 November 2014 at 06:17, Alshikho, Mohamad J.
> > <[log in to unmask]<mailto:[log in to unmask]>> wrote:
> >>
> >> Dear Anderson,
> >> I ran the TBSS analysis for two groups 23 patients and 27 controls. I
> >> collected the raw FA images for all the subjects in one folder then I
> >> organized the data in this order (controls then patients).
> >> I did the following:
> >> Tbss_1_ preproc *.nii.gz
> >> Tbss_2_reg -T
> >> Tbss_3_postreg -S (I designed template from the best 24 FA images that I
> >> have )
> >> Tbss_4_prestats 0.2
> >> design_ttest2 design 27 23
> >> randomise -i all_RD_skeletonised -o tbss_RD -m mean_FA_skeleton_mask -d
> >> design.mat -t design.con -n 10000 --T2 -V (I used 1000 permutations)
> >>
> >> Then I ran the analysis again for the non_FA maps.
> >> At the end I ran the cluster command line as the following:
> >>
> >> cluster -i tbss_tfce_corrp_tstat2.nii.gz -t 0.95 -o cluster_index --mm
> >> --osize=cluster_size > cluster_info.txt
> >> cluster -i tbss_MD_tfce_corrp_tstat2.nii.gz -t 0.95 -o cluster_index
> --mm
> >> --osize=cluster_size > cluster_info_MD.txt
> >> cluster -i tbss_AD_tfce_corrp_tstat2.nii.gz -t 0.95 -o cluster_index
> --mm
> >> --osize=cluster_size > cluster_info_AD.txt
> >> cluster -i tbss_RD_tfce_corrp_tstat2.nii.gz -t 0.95 -o cluster_index
> --mm
> >> --osize=cluster_size > cluster_info_RD.txt
> >>
> >> For all the previous commands (cluster ) I got one big cluster.
> >>
> >> The previous steps are exactly what I did. Looking forward for your
> >> support!
> >>
> >> Thanks a lot
> >> Mohamad
> >>
> >>
> >>
> >>
> >>
> >>
> >> From: FSL - FMRIB's Software Library [[log in to unmask]<mailto:
> [log in to unmask]>] on behalf of
> >> Anderson M. Winkler [[log in to unmask]<mailto:
> [log in to unmask]>]
> >> Sent: Saturday, November 29, 2014 2:03 PM
> >>
> >> To: [log in to unmask]<mailto:[log in to unmask]>
> >> Subject: Re: [FSL] cluster & atlasquery
> >>
> >> Hi Mohamad,
> >>
> >> It's not possible to tell only with the information you provided, but
> >> assuming that everything went fine, and that design and contrasts are
> fine,
> >> it can be that there is a global effect.
> >>
> >> However, the output of "cluster" shows that your maximum in the image is
> >> 1, but this isn't in the outputs fromn randomise, because this max
> >> represents a 1-p, and the minimum p is limited by the number of
> >> permutations. In other words, if you run 5000 permutations, the max
> should
> >> be, at best, 0.9998, not 1. Please, check if you didn't do other
> processing
> >> in these outputs that shouldn't have been done.
> >>
> >> All the best,
> >>
> >> Anderson
> >>
> >>
> >> On 28 November 2014 at 15:55, Alshikho, Mohamad J.
> >> <[log in to unmask]<mailto:[log in to unmask]>> wrote:
> >>>
> >>> Dear Anderson,
> >>> thank you very much for your answer.
> >>> Kindly what is/ are the reasons for this result (1 big cluster with
> many
> >>> tracts). Is it because of a technical issue or it is related to the
> data
> >>> itself.
> >>> Regarding the coordinate : I tried to find the coordinates manually
> using
> >>> Fslview then I used
> >>> fslmeants -i all_FA_skeletonized -c x y z
> >>> to extract FA value for every subject tract
> >>>
> >>> Is this correct or do you suggest me better Idea.
> >>>
> >>> I highly appreciate your help in this
> >>> Thanks
> >>> ________________________________
> >>> From: FSL - FMRIB's Software Library [[log in to unmask]<mailto:
> [log in to unmask]>] on behalf of
> >>> Anderson M. Winkler [[log in to unmask]<mailto:
> [log in to unmask]>]
> >>> Sent: Friday, November 28, 2014 10:12 AM
> >>> To: [log in to unmask]<mailto:[log in to unmask]>
> >>> Subject: Re: [FSL] cluster & atlasquery
> >>>
> >>> Hi,
> >>> You have just 1 big cluster with many tracts inside it. I don't think
> >>> finding coordinates will be helpful, sorry...
> >>> All the best,
> >>> Anderson
> >>>
> >>>
> >>>
> >>> On 27 November 2014 at 14:54, M. alshikho <[log in to unmask]
> <mailto:[log in to unmask]>>
> >>> wrote:
> >>>>
> >>>> Dear FSL experts,
> >>>> In order to report the results of TBSS analysis, I did the following:
> >>>>
> >>>> cluster -i tbss_RD_tfce_corrp_tstat2.nii.gz -t 0.95 -o cluster_index
> >>>> --mm --osize=cluster_size > cluster_info.txt
> >>>>
> >>>> The content of cluster_info is
> >>>> Cluster Index Voxels MAX MAX X (mm) MAX Y (mm) MAX Z
> >>>> (mm) COG X (mm) COG Y (mm) COG Z (mm)
> >>>> 1 41305 1 11 -20 -1 0.979 -26.5 2.56
> >>>>
> >>>> Then I used fslmaths -dt int cluster_index.nii.gz -thr 1 -uthr 1 -bin
> >>>> cluster_mask1 to create mask for this cluster.
> >>>> Finaly I used atlas query as the following:
> >>>> atlasquery -a "JHU White-Matter Tractography Atlas" -m
> >>>> cluster_mask1.nii.gz
> >>>>
> >>>> the results are as the following:
> >>>> Anterior thalamic radiation L:1.77475
> >>>> Anterior thalamic radiation R:1.53817
> >>>> Corticospinal tract L:1.3281
> >>>> Corticospinal tract R:1.41649
> >>>> Cingulum (cingulate gyrus) L:0.240552
> >>>> Cingulum (cingulate gyrus) R:0.00225154
> >>>> Cingulum (hippocampus) L:0.297494
> >>>> Cingulum (hippocampus) R:0.00769883
> >>>> Forceps major:0.383101
> >>>> Forceps minor:0.0229028
> >>>> Inferior fronto-occipital fasciculus L:1.04522
> >>>> Inferior fronto-occipital fasciculus R:1.22639
> >>>> Inferior longitudinal fasciculus L:1.68708
> >>>> Inferior longitudinal fasciculus R:1.10311
> >>>> Superior longitudinal fasciculus L:1.63842
> >>>> Superior longitudinal fasciculus R:2.11788
> >>>> Uncinate fasciculus L:0.236122
> >>>> Uncinate fasciculus R:0.013945
> >>>> Superior longitudinal fasciculus (temporal part) L:0.725336
> >>>> Superior longitudinal fasciculus (temporal part) R:0.91892
> >>>>
> >>>> Now I have one cluster and the probability for many tracts in the same
> >>>> cluster. How can I define the coordinate for every tracts from the
> output of
> >>>> atalsquery?
> >>>>
> >>>> Thanks
> >>>
> >>>
> >>> The information in this e-mail is intended only for the person to whom
> it
> >>> is
> >>> addressed. If you believe this e-mail was sent to you in error and the
> >>> e-mail
> >>> contains patient information, please contact the Partners Compliance
> >>> HelpLine at
> >>> http://www.partners.org/complianceline . If the e-mail was sent to
> you in
> >>> error
> >>> but does not contain patient information, please contact the sender and
> >>> properly
> >>> dispose of the e-mail.
> >>
> >>
> >
>
>
>
> --
> Niels Bergsland
> Integration Director
> Buffalo Neuroimaging Analysis Center
> 100 High St. Buffalo NY 14203
> [log in to unmask]<mailto:[log in to unmask]>
>
